基于SIFT特征的哈希快速检索与图像匹配
|
摘要
针对SIFT算法在应用于图像匹配时,存在准确率低下和耗时等问题,提出一种SIFT特征的哈希快速检索与图像匹配方法。文中提出以二值化SIFT关键点描述子和哈希表相结合的方法对图像进行匹配。针对实验过程中出现的冲突项,通过在哈希表中添加标志位并记录冲突相个数和地址,完美地解决了高维描述子转化到低维冲突项的问题,加快了匹配速度。实验结果表明,该方法图像匹配速度优于传统SIFT匹配方法,加快了相似特征检索速度、提高了查询效率,并能够满足实时应用。所提出的采用SIFT关键点描述子的二值化与哈希检索相结合的方法,通过对比实验,证明了该方法在保证准确率的同时,提高了效率,实现了图像的实时快速匹配。
In allusion to the low accuracy rate and time-consumption problems existing during the application of the SIFT algorithm in image matching,a hash fast retrieval and image matching method based on the SIFT feature is proposed. A method combining the binarized SIFT key point descriptors and hash table is proposed to conduct image matching. For the conflict items that appear in the experiment,the problem of transforming the high-dimensional descriptors to low-dimensional conflict items is perfectly solved and the matching speed is accelerated by adding the flag and recording the number and addresses of conflicts in the hash table. The experimental results show that in comparison with the traditional SIFT matching method,the method has a better image matching speed,can accelerate the retrieval speed of similar features,improve the query efficiency and meet the real-time application requirement. The proposed method combining the binarization of SIFT key point descriptors and hash retrieval can improve the efficiency and realize real-time and fast matching of images at the prerequisite of ensuring the accuracy,which is demonstrated in the comparison experiment.
In allusion to the low accuracy rate and time-consumption problems existing during the application of the SIFT algorithm in image matching,a hash fast retrieval and image matching method based on the SIFT feature is proposed. A method combining the binarized SIFT key point descriptors and hash table is proposed to conduct image matching. For the conflict items that appear in the experiment,the problem of transforming the high-dimensional descriptors to low-dimensional conflict items is perfectly solved and the matching speed is accelerated by adding the flag and recording the number and addresses of conflicts in the hash table. The experimental results show that in comparison with the traditional SIFT matching method,the method has a better image matching speed,can accelerate the retrieval speed of similar features,improve the query efficiency and meet the real-time application requirement. The proposed method combining the binarization of SIFT key point descriptors and hash retrieval can improve the efficiency and realize real-time and fast matching of images at the prerequisite of ensuring the accuracy,which is demonstrated in the comparison experiment.
引文
[1]王凡.基于SIFT的图像检索特征改进方法[J].数字技术与应用,2016(1):139-141.WANG Fan. Image retrieval feature improvement method based on SIFT[J]. Numerical technology&applications,2016(1):139-141.
[2]方壮.一种迭代有序K最邻近距离实现数字图像特征点匹配的算法[J].湖北民族学院学报(自然科学版),2013,31(1):36-37.FANG Zhuang. A kind of digital image feature points matching algorithm realized by sequential iteration K-neighbor[J]. Journal of Hubei University for Nationalities(Natural science edition),2013,31(1):36-37.
[3]闫家梅.基于SIFT特征的人脸识别[J].科学技术与工程,2013,13(5):1219-1222.YAN Jiamei. Face recognition based on SIFT features[J]. Science technology and engineering,2013,13(5):1219-1222.
[4]刘兆庆,李琼,刘景瑞,等.一种基于SIFT的图像哈希算法[J].仪器仪表学报,2011,32(9):2024-2028.LIU Zhaoqing,LI Qiong,LIU Jingrui,et al. SIFT based image hashing algorithm[J]. Chinese journal of scientific instrument,2011,32(9):2024-2028.
[5]胡鹏.基于嵌入式系统的自适应SIFT图像配准算法的研究[D].西安:西安电子科技大学,2014.HU Peng. Embedded system based image registration using adaptive SIFT algorithm[D]. Xi’an:Xidian University,2014.
[6]黄东晓.基于三线阵CCD的新型三维形貌测量系统研究[D].天津:天津大学,2014.HUANG Dongxiao. Study of a new 3D shape measurement system based on tri-linear CCDs[D]. Tianjin:Tianjin University,2014.
[7]刘俊.基于多特征融合的奶牛图像识别系统研究[D].上海:上海师范大学,2013.LIU Jun. The research on cow image recognition system based on multi-feature fusion[D]. Shanghai:Shanghai Normal University,2013.
[8]李学超.基于斑点跟踪算法的心脏超声图像运动分析[D].沈阳:东北大学,2013.LI Xuechao. Motion analysis on cardiac ultrasound images based on speckle tracking technology[D]. Shenyang:Northeastern University,2013.
[9] CHEN C C,HSIEH S L. Using binarization and hashing for efficient SIFT matching[D]. Journal of visual communication and image presentation,2015,30:86-93.
[10] HE T,WEI Y,LIU Z,et al. Content based image retrieval method based on SIFT feature[C]//Proceedings of 2018 International Conference on Intelligent Transportation,Big Data&Smart City. Ji’nan:IEEE Computer Society,2018:649-652.
[2]方壮.一种迭代有序K最邻近距离实现数字图像特征点匹配的算法[J].湖北民族学院学报(自然科学版),2013,31(1):36-37.FANG Zhuang. A kind of digital image feature points matching algorithm realized by sequential iteration K-neighbor[J]. Journal of Hubei University for Nationalities(Natural science edition),2013,31(1):36-37.
[3]闫家梅.基于SIFT特征的人脸识别[J].科学技术与工程,2013,13(5):1219-1222.YAN Jiamei. Face recognition based on SIFT features[J]. Science technology and engineering,2013,13(5):1219-1222.
[4]刘兆庆,李琼,刘景瑞,等.一种基于SIFT的图像哈希算法[J].仪器仪表学报,2011,32(9):2024-2028.LIU Zhaoqing,LI Qiong,LIU Jingrui,et al. SIFT based image hashing algorithm[J]. Chinese journal of scientific instrument,2011,32(9):2024-2028.
[5]胡鹏.基于嵌入式系统的自适应SIFT图像配准算法的研究[D].西安:西安电子科技大学,2014.HU Peng. Embedded system based image registration using adaptive SIFT algorithm[D]. Xi’an:Xidian University,2014.
[6]黄东晓.基于三线阵CCD的新型三维形貌测量系统研究[D].天津:天津大学,2014.HUANG Dongxiao. Study of a new 3D shape measurement system based on tri-linear CCDs[D]. Tianjin:Tianjin University,2014.
[7]刘俊.基于多特征融合的奶牛图像识别系统研究[D].上海:上海师范大学,2013.LIU Jun. The research on cow image recognition system based on multi-feature fusion[D]. Shanghai:Shanghai Normal University,2013.
[8]李学超.基于斑点跟踪算法的心脏超声图像运动分析[D].沈阳:东北大学,2013.LI Xuechao. Motion analysis on cardiac ultrasound images based on speckle tracking technology[D]. Shenyang:Northeastern University,2013.
[9] CHEN C C,HSIEH S L. Using binarization and hashing for efficient SIFT matching[D]. Journal of visual communication and image presentation,2015,30:86-93.
[10] HE T,WEI Y,LIU Z,et al. Content based image retrieval method based on SIFT feature[C]//Proceedings of 2018 International Conference on Intelligent Transportation,Big Data&Smart City. Ji’nan:IEEE Computer Society,2018:649-652.